A Texas jury strikes down a police program that used AI and a rigged dog to search cars at random

By Reason Magazine · September 30, 2026.

A federal jury on Monday found Bexar County, Texas, liable for systematically violating the Fourth Amendment —which protects against unreasonable searches and seizures— for years, through a "criminal interdiction" unit that combined automatic license plate readers with artificial intelligence and traffic stops based on fabricated pretexts. The case was brought by Alek Schott, a driver pulled over in March 2022 near San Antonio for allegedly drifting out of his lane, something his own dashcam later disproved, according to the lawsuit filed by the Institute for Justice, the nonprofit that represented him.

What was meant to be a routine stop turned into an hour of questioning. Officer Joel Babb had Schott sit in the passenger seat of the patrol car and asked him about his job, his address and his record while he checked it, according to the lawsuit. Ten minutes in, Babb admitted the real reason was not any traffic matter: he belonged to the Criminal Interdiction Unit and his job was to look for human and drug smuggling, not to write tickets.

When Schott refused to let them search the vehicle, Babb called in a K-9 unit. Body camera video shows the officer making upward gestures next to the dog just before it put its paws on the car; that gesture was recorded as a "positive alert" which, legally, authorizes a search without consent. Over the next 40 minutes, the officers went through Schott's truck. They found nothing. As he said goodbye, Babb confessed to him that "nine times out of ten" the same thing happens: they find nothing.

According to the lawsuit, that pattern was not one officer's isolated excess, but a five-step system: a network of AI-equipped license plate readers flagged driving patterns labeled "anomalous"; Bexar officers used that signal to justify a stop on a traffic pretext; they questioned the driver about drugs, cash and consent to search; if the driver refused, an alert from a dog cued by the officer supplied the missing "probable cause". The Institute for Justice contends that the unit subjected "thousands" of drivers to this routine. "It rests on the idea that if you watch enough innocent people closely, criminals will eventually turn up," the organization's senior attorney, Josh Windham, said in a statement.

The lead attorney on the case, Christie Hebert, went further at the press conference after the verdict: officers need genuine suspicion of a crime to stop, detain and search someone, and in this case they did not have it. "Every agency in the state of Texas should know that we are ready to enforce this verdict," she warned. The jury awarded Schott $76 —one dollar for every minute he was detained without cause—, a symbolic figure that contrasts with the ruling's real reach: a declaration that the program, as it operated, was unconstitutional.

As industry context, this type of automatic license plate reader (ALPR) network has expanded sharply in the United States in recent years through private vendors that connect cameras from thousands of local agencies into a single national database, selling the promise of detecting "anomalous behavior" in real time. It is an infrastructure designed for public safety that, without strict judicial oversight, can turn any driver into a suspect by default.

Here is the underlying problem, and it is the one we are most interested in flagging: AI did not invent police abuse, but it did give it industrial scale. Before, an officer needed a suspicion in order to notice a particular car; now an algorithm generates the suspicion for him, from the mobility patterns of thousands of people who have done nothing. What on paper is presented as "efficiency" —directing limited resources toward the most likely traffic— translates, without rigorous human oversight, into mass surveillance disguised as a neutral technical tool. The human component of this story, the gesture toward the dog to force a false alert, is the reminder that automating suspicion does not eliminate bad faith: it only gives it a layer of apparent algorithmic objectivity to hide behind.

Our reading is that this verdict matters less for the $76 than for the signal it sends: the courts, and ultimately a jury of ordinary citizens, remain the brake that works when surveillance technology is deployed faster than the oversight meant to accompany it. It is exactly the kind of short-term friction that our underlying thesis acknowledges bluntly —AI applied without institutional checks erodes concrete rights, of concrete people, right now— and it is not resolved with generic techno-optimism. It is resolved with lawsuits like this one, with juries willing to say that a "tool" does not exempt anyone from complying with the Fourth Amendment, and with clear rules on what can and cannot be automated in the state's use of force. The promise of an AI that frees up resources and time for people only holds if, in parallel, we safeguard against that same capability becoming a tool of disproportionate control over those who have done nothing wrong. This case is a one-off victory in that direction, and the Institute for Justice has already warned that it is prepared to enforce this verdict against any agency in the state of Texas.